Data handling and record keeping
Each participant pair had a unique identification (ID) number. All data recorded on individuals used the anonymous ID. All forms with subject names is kept in a locked cabinet, when not in use. Clinical data is stored separately from that containing personal information. Data will be stored for at least 10 years. The informed consent forms and source documents will be stored, after study completion, in a confidential archive; they will be stored separately from the study documents in the access-restricted archive at the MRC Unit.
Some sections of the CRFs act as source data for the trial and are completed directly by the relevant research personnel. A few sections, such as vaccination, are transcribed from the Infant Welfare Card for the newborn. Other information from the mother and the baby (such as age of the mother or birth weight of the baby) was transcribed from the antenatal card. Other source data include: the structured clinical notes at enrollment, discharge and post-natal visit, and the structured information collected from field workers/nurses during the weekly home visits (week 2 to week 8). At each visit, a CRF was completed by the relevant study team member. In addition, laboratory forms were also completed for any sample collected as part of the study from either the mother or the baby.
CRF data is entered into an
OpenClinica (
www.openclinica.com) database using double data entry and verification. Laboratory data have been recorded on lab forms designed specifically for the study and entered into a separate site within the
OpenClinica study database. Screening information was entered in another sub-database within the main
OpenClinica database. Screening records are uniquely identified by the sensitization number, which is issued at the time of sensitization. Consistency checks have been performed during data entry and outliers and missing data checked against the original forms then subsequently amended in the database.
The PI and sponsor maintain appropriate medical and research records for this study in compliance with applicable Good Clinical Practice (GCP), regulatory and institutional requirements. Authorised representatives of the sponsor, the ethics committee or regulatory bodies may inspect all documents and records maintained by the investigator. The PI or designee ensures access to facilities and the records.
The PI or designee documented and explained any deviation from the approved protocol and reported them to the sponsor and ethics committee in accordance with the MRC Unit’s standard procedures.
At study closure, all items on the following checklist will be checked prior to locking the database:
-
All expected study subjects have been entered
-
Checks for any duplicate subjects is negative
-
SAE reconciliation has been completed
-
Protocol Deviation reconciliation has been completed.
-
All final cleaning checks have been performed and all outstanding issues resolved
-
Approval for database lock has been obtained from the Head of Data Management (HoDM), the statistician, and the PI.
-
Database access rights have been removed.
The randomisation list will be kept by the statistician of the DSMB until the blinded results of the statistical analysis are available. The study site will retain a set of envelopes containing the allocation of each subject so that individuals can be unblinded for safety reasons without accessing the randomization list.
Trial monitoring
A clinical trials monitor independent from the investigator team monitored the trial. The monitor conducted an initiation visit to the site before the trial started to verify and document that all study staff were adequately informed about the study and that the IMP to be used had been received and stored accordingly on site.
During the course of the trial, interim monitoring visits took place at the health centre, the IMP storage facility, the site office and the laboratories.
During the visits, the monitor carried out a quality control of trial progress in respect to protocol and operating guidelines, data collection, completion of consent forms, completion of study documents, SAE reporting, and samples and IMP management. Source data verification was conducted for all ICFs, eligibility criteria and SAEs.
The monitor inspected up to 10 % of participants CRFs for accuracy, completeness and consistency with source documents before data entry and reviewed the investigator’s file.
At the end of each monitoring visit, the monitor discussed the findings with the study team and outlined what further actions should be taken.
The Monitor will conduct a close-out visit after all data queries have been resolved and the data base has been locked to ensure that all essential documents are completed and filed accordingly and are ready for archiving, that all IMPs have been accounted for and are returned to supplier or destroyed accordingly, and that all unused study materials are returned to suppliers or destroyed, as applicable.
The study team established, in addition, internal monitoring procedures to ensure review of all CRFs by study team members before any data was entered into the database.
Statistical analysis
A flow chart will be used to describe the number of mothers who were eligible, randomised and treated, and the number of samples obtained from mothers and children at each time point.
The adequacy of the randomisation procedure will be checked by comparing baseline characteristics of mothers (ethnicity, age, season enrolled, hours from treatment to delivery, mode of delivery, time to rupture of membrane, apgar score, multiple pregnancy) and newborns (sex, birth weight, gestational age) between arms. Some of these variables are measured post intervention, but are unlikely to be influenced by the intervention (time from treatment to delivery, mode of delivery, time to rupture of membrane, Apgar score). Prevalence of bacterial carriage in the mother pre-intervention will also be used to check comparability of baseline characteristics (NPS and vaginal swab).
The primary analysis will be based on intention to treat (ITT). We will include data on twins but will not adjust for the effect of clustering (the average cluster size will be close to 1 so the design effect will be negligible). A per protocol analysis will be conducted excluding samples from mothers who vomited less than 15 min after medication was taken, samples collected after the use of antibiotics, and samples arriving at the lab more than 8 h after collection.
The prevalence of bacterial carriage will be compared between arms at each time point (in NPS, vaginal swabs and breast milk). The prevalence of carriage will be presented separately for each species and combined (i.e., carriage of S.aureus, GBS or S.pneumoniae). In addition, carriage acquisition rates will be compared between arms for the periods 0–6 days and 7–28 days (acquisition will happen when a non-carrier will become carrier in the subseuqent sample). Ratios of prevalence and 95 % confidence intervals will be calculated for each of these comparisons, and Fisher’s exact test will be used to compute p-values for the comparisons of prevalence.
A Poisson regression model will be used to assess the effect of the intervention on prevalence of carriage across all time points. The model will include time, baseline carriage status of the mother, mother’s age, ethnicity, season, sex of the newborn, and gestational age. Robust standard errors will be used to account for the dependence between observations from the same individual as well as the model misspecification (carriage status is binary and therefore does not follow a Poisson distribution).
To calculate prevalence of carriage at day 3 and day 6 we will use data from samples (NPS newborns and mothers, breast milk) collected within 1 day of the scheduled visit. At 14 days we will use samples collected between 12 and 18 days, and at 28 days we will use samples collected between 25 and 31 days. For vaginal bacterial carriage at day 8–10 we will use samples collected between 7 and 13 days. Samples were not collected after children received antibiotics. Where a sample was collected in lieu of the missing sample (i.e., before antibiotics were given) this will be used in place of the missing sample, even if it was collected outside the appropriate window period.
We will present analyses based on the available data. For the primary outcome we will conduct two additional analyses using the ITT cohort. First, a subgroup analysis in women who gave birth 2 h or more after taking the drug. And second, a sensitivity analysis using multiple imputation of missing data. In the imputation model, bacterial carriage in missing samples will be imputed from data on bacterial carriage at other time points and baseline demographic data (age and ethnicity for mothers, and sex and birth weight for newborns).
Changes in prevalence of bacterial carriage over time will be plotted for NPS in mothers and newborns, and breast milk. The plots will include point-wise 95 % confidence intervals for each prevalence estimate.
The numbers and rates of adverse events, serious adverse events (overall and specifically with a diagnosis of neonatal sepsis), purulent conjunctivitis, Chlamydial conjunctivitis, and deaths in newborns will be reported by trial arm. Rates of adverse events and serious adverse events will be reported in mothers. Rates will be compared between trial arms using Poisson regression, and robust standard errors will be used to allow for heterogeneity in the rates between individuals.